Electromechanical equipment state digital twin driving model based on multi-sensor monitoring

Abstract

The invention relates to the technical field of industrial digital twinning manufacturing, and discloses an electromechanical equipment state digital twinning driving model based on multi-sensor monitoring, and the model comprises the steps: collecting the original vibration signal data of mechanical equipment through a vibration sensor, and transmitting the original vibration signal data to edge computing equipment for coding; performing clustering on a normal point signal coefficient and an outlier signal coefficient obtained after discrete cosine transform by using a 3 sigma criterion; quantizing a normal point signal coefficient and an outlier signal coefficient corresponding to each vibration sensor by using a selected bit number, then fusing the normal point signal coefficient and the outlier signal coefficient, and carrying out lossless compression on the obtained fusion coefficient; and transmitting the compressed data from the edge computing device to a twin virtual side for decoding, unpacking fusion and quantization inversion. According to the method, the data volume is reduced through discrete cosine transform processing of original data, clustering of discrete cosine transform coefficients, quantitative mapping and coefficient fusion, and the data reconstruction precision is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial digital twin manufacturing, and in particular to a digital twin driving model of electromechanical equipment status based on multi-sensor monitoring. Background Art

[0002] Digital Twin (DT) is a new digital factory technology that provides dynamic acquisition of equipment status in industrial environments. It builds a digital model of the corresponding physical entity in the virtual space through advanced digital technologies such as software. At the same time, the Internet of Things technology is used to transmit digital twin data in real time, and the physical status of the equipment is mapped to the virtual model in real time, so that the digital space can respond to the state changes in the physical space. In the industrial production environment, mechanical equipment plays an important role and is an indispensable basic physical equipment subject for building digital twins. In a modern smart factory based on digital twins, the operating status of machinery and equipment needs to be presented in the digital twin. Many physical parameters can reflect the operating status of mechanical equipment, such as temperature, speed, stress, strain and vibration. Among them, vibration signals can accurately reflect the failure of equipment, and the technology for obtaining vibration signals of mechanical equipment is relatively mature. In addition, mechanical vibration is a non-hysteresis variable, which is more real-time than other slowly changing state quantities and more sensitive to equipment response.

[0003] In actual industrial environments, the amount of data on mechanical equipment and measurement points that need to be measured is huge, and the real-time performance of the data is highly required. For example, a single 32-bit sensor represents a sampling point at a sampling frequency of 50kHz, generating 200kB of data per second. In mechanical digital twin systems, the amount of measurement point data usually grows exponentially, and data transmission is prone to congestion when the network bandwidth is narrow. Therefore, the vibration signal data needs to be compressed on the side close to the physical entity to reduce the amount of reconstructed data on the virtual side of the twin.

[0004] Existing digital twin mechanical vibration data encoding and decoding methods are almost all for a single sensor, which has problems such as low compression rate, poor compression timeliness, low continuity, and large bandwidth occupancy. Therefore, by compressing and encoding the data of multi-sensor digital twins, the amount of data transmission can be greatly reduced while maintaining a high reconstruction accuracy. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides a digital twin driving model of electromechanical equipment status based on multi-sensor monitoring. First, the original mechanical vibration signals are collected from multiple front-end sensors and transmitted to the edge computing entity device for encoding; then the original vibration signal data is discrete cosine transformed, and the normal point signal coefficients and outlier signal coefficients after the discrete cosine transform of the original vibration signal data are clustered using the 3-sigma criterion; the outlier signal coefficients of each sensor are quantized with an appropriate number of bits, and then the quantized normal point signal coefficients and outlier signal coefficients are fused in bytes respectively, and the fusion coefficients are losslessly compressed; finally, the compressed data is transmitted from the edge computing device to the twin virtual side for decoding, unpacking, fusion and quantization inversion, thereby realizing data reconstruction of the twin virtual side of the mechanical equipment.

[0006] The present invention specifically adopts the following technical solutions:

[0007] A digital twin driving model of electromechanical equipment status based on multi-sensor monitoring, the method adopted includes the following steps:

[0008] Step S1, using a vibration sensor to collect original vibration signal data of mechanical equipment, and transmitting the original vibration signal data to an edge computing device for encoding;

[0009] Step S2, performing discrete cosine transform on the encoded original vibration signal data, and clustering the normal point signal coefficients and outlier signal coefficients obtained after the discrete cosine transform using the 3-ξ criterion;

[0010] Step S3, quantizing the normal point signal coefficient and the outlier signal coefficient corresponding to each vibration sensor using a selected number of bits, then fusing the quantized normal point signal coefficient and the outlier signal coefficient in bytes, and performing lossless compression on the obtained fused coefficients;

[0011] Step S4, transmits the compressed data from the edge computing device to the twin virtual side for decoding, unpacking, fusion and quantization inversion to obtain a reconstructed signal, thereby realizing data reconstruction of the twin virtual side of the mechanical equipment.

[0012] Furthermore, step S1 specifically includes:

[0013] Multiple vibration sensors are placed around the planetary gearbox of the mechanical equipment to collect the original vibration signal data of the mechanical equipment. Each vibration sensor is responsible for collecting data from a vibration test point. The data sampling card is connected to the edge computing device for fusion encoding of the vibration signal.

[0014] Furthermore, step S2 specifically includes the following steps:

[0015] S21, define the original vibration signal data X:

[0016]

[0017] Where m is the number of vibration sensors on the mechanical equipment, and n is the number of sampling points of each vibration sensor in a period of time;

[0018] Perform discrete cosine transform on the original vibration signal data X to obtain the discrete cosine transform coefficient matrix C:

[0019]

[0020] Among them, the discrete cosine transform coefficient k=1,2,...,n;

[0021] S22, discrete cosine transform coefficients represent the degree of signal energy concentration. Larger amplitudes are close to zero, and smaller amplitudes are far from zero. Discrete cosine transform coefficients with amplitudes greater than a set value and amplitudes less than a set value are defined as outlier coefficients and normal point coefficients, respectively.

[0022] S23, outlier coefficient detection: First, the elements in the discrete cosine transform coefficient matrix of each vibration sensor are arranged from the smallest to the largest into a sequence matrix C o :

[0023]

[0024] Represents the sequence matrix C o The elements in the distribution are calculated using the Shapiro-Wilk test method. i :

[0025]

[0026] Among them, when ξi>ξ(α), the distribution test value ξi is determined to obey the normal distribution, ξ(α) is obtained by querying the Shapiro-Wilk test table, α=0.05, n is the number of discrete cosine transform coefficients, is the Shapiro-Wilk test table coefficient, μ i is the discrete cosine transform coefficient c ij The average value of:

[0027]

[0028] Calculate the distribution test value using Kolmogorov-Smirnov

[0029]

[0030] in, When , the distribution test value It is judged to be subject to normal distribution. From the Kolmogorov-Smirnov test table, we can find that α = 0.05, where F(c ij ) is c ij The cumulative probability density function of for The cumulative frequency of F(c ij )and The calculation formula is as follows:

[0031] Formula 7:

[0032]

[0033] In the formula, It's about c ij The probability density function of is:

[0034]

[0035] Among them, σ i is the standard deviation of the discrete cosine transform coefficient of the i-th vibration sensor:

[0036]

[0037] Where, j is the subscript sequence of the i-th vibration sensor;

[0038] The 3-ξ criterion is used to test the normal distribution of discrete cosine transform coefficients and then screen outliers:

[0039] P{|c ii -μ i |>(ξ*σ i )}=P(ξ);

[0040] Among them, P(ξ=3)=0.9973, if |c ij -μ i |greater than ξ*σ i , then P{|c ij -μ i |>(ξ*σ i )} can be calculated by formula 7; the larger the value of ξ, the larger the threshold for screening outliers;

[0041] S24, in the encoding process of the original vibration signal data, the normal point coefficients and the outlier coefficients are quantized and stored respectively. When encoding the normal point coefficients, the normal point coefficients of specific bits are quantized. When encoding the outlier coefficients, in addition to quantizing the specific values ​​of the outlier coefficients, the order corresponding to each outlier coefficient is also recorded and stored separately.

[0042] Furthermore, step S3 specifically includes the following steps:

[0043] S31, Discrete cosine transform coefficient bit quantization number:

[0044]

[0045] Among them, χ max and χ min are the maximum and minimum values ​​of the interval to which each discrete cosine transform coefficient is mapped. imax and c imin is the maximum and minimum value of the coefficient to be quantized, is the quantization coefficient of any sensor, χ max and χ min The calculation is as follows:

[0046] χ max =2 Q -1, χ min =0;

[0047] Among them, Q is the number of bits of discrete cosine transform coefficient quantization, and the inverse quantization change is expressed as:

[0048]

[0049] in, is the inverse quantization fusion coefficient; and are the maximum and minimum values ​​of the quantization coefficients respectively;

[0050] S32, Discrete Cosine Transform Coefficient Fusion: Byte fusion is used to mix the discrete cosine transform coefficients of multiple vibration sensors into a fusion file; discrete cosine transform coefficient fusion includes outlier coefficient fusion and normal point coefficient fusion:

[0051]

[0052] in, In the fusion file The probability of, H is the information entropy of the fusion file; for m vibration sensors, n points are taken for discrete cosine transform, and there are still n coefficients after the transformation. After filtering the discrete cosine transform coefficients of each sensor, the outlier coefficients are fused with the normal point coefficients respectively;

[0053] Fusion sequence of normal point coefficients and the fusion sequence of the outlier coefficient It is expressed as:

[0054]

[0055]

[0056] Among them, the set of normal point coefficients is S = {s|(η, κ, ..., v)}; the set of outlier coefficients is o = ns∈S, η, κ, v∈S≤n, and Respectively represent the normal point quantization coefficient and the outlier point quantization coefficient;

[0057] Bit fusion is used to remove the information redundancy of the fusion sequence of normal point quantization coefficients and outlier point quantization coefficients. First, the fusion sequence and Each coefficient is extracted as an 8-bit unsigned integer byte, each byte represents only 0 or 1:

[0058]

[0059] in is the byte to be extracted, “>>” is the symbol for right shift, “&” is the symbol for AND operation, k is the number of right shifts, and q is the quantization bit;

[0060] After extracting the fused sequence as bytes, bit fusion is performed as follows:

[0061]

[0062] in is a new byte fused with the first 8 bits, “<<” is the symbol for left shift, “|” is the symbol for the “or” operation, and k is the number of left shifts.

[0063] Furthermore, step S4 specifically includes the following steps:

[0064] S41, data decoding: decompressing the received compressed data and reading the header information of the decompressed data file;

[0065] S42, signal reconstruction: including coefficient quantization, bit fusion and lossless compression; the number of data compression bits B is:

[0066]

[0067] Among them, N sv，i is the specific value of the number of bits of the i-th sensor, B rv and B ov are the specific values ​​of the quantization bit number of the normal point discrete cosine transform coefficient and the quantization bit number of the outlier discrete cosine transform coefficient respectively; N ri and N oi are the specific values ​​of the number of fused unquantized normal point discrete cosine transform coefficients and the number of fused unquantized outlier discrete cosine transform coefficients of the i-th sensor respectively;

[0068] The reconstructed signal is calculated as follows:

[0069]

[0070] in, is the inverse quantization fusion coefficient, is the inverse discrete cosine transformed vibration data.

[0071] Compared with the prior art, the beneficial technical effects of the present invention are:

[0072] The present invention reduces the amount of data by discrete cosine transform processing of original data, clustering of discrete cosine transform coefficients, quantization mapping and coefficient fusion, while ensuring data reconstruction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A schematic diagram of a method flow chart provided according to an embodiment of the present invention;

[0074] Figure 2 This is a diagram of a signal reconstruction result provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0075] A preferred embodiment of the present invention is described in detail below with reference to the accompanying drawings.

[0076] The method adopted by the digital twin driving model of electromechanical equipment state based on multi-sensor monitoring in the present invention includes the following steps:

[0077] Step S1: Use a vibration sensor to collect original vibration signal data of mechanical equipment, and transmit the original vibration signal data to an edge computing device for encoding.

[0078] Step S1 specifically includes:

[0079] Eight vibration sensors are placed around the planetary gearbox to collect vibration signals of mechanical equipment. Each vibration sensor is responsible for collecting one vibration test point. The data sampling card is connected to the edge computing device for vibration signal fusion encoding. The sampling frequency is set to 128000Hz to collect the original vibration signal.

[0080] Step S2: Perform discrete cosine transform on the encoded original vibration signal data, and cluster the normal signal coefficients and outlier signal coefficients after discrete cosine transform of the original vibration signal data using the 3σ criterion.

[0081] Step S2 specifically includes:

[0082] S21, define the original vibration signal data X as:

[0083]

[0084] Among them, m is the number of mechanical vibration sensors collected simultaneously, and n is the number of points sampled by each sensor in a period of time.

[0085] The discrete cosine transform is applied to the original vibration signal data X, and the discrete cosine transform coefficient matrix C can be expressed as:

[0086]

[0087] in,

[0088] S22, discrete cosine transform coefficients represent the degree of signal energy concentration. Larger amplitudes are close to zero, and smaller amplitudes are far away from zero. The coefficients with larger amplitudes and smaller amplitudes are defined as outlier coefficients and normal point coefficients, respectively.

[0089] S23, outlier coefficient detection: The outlier discrete cosine transform coefficient screening based on statistics requires determining whether all discrete cosine transform coefficients conform to the normal distribution. The Shapiro-Wilk test (Shapiro-Wilk, SW) and Kolmogorov-Smirnov (Kolmogorov-Smirnov, KS) are used for normality tests when the number of coefficients to be tested is less than 50 and greater than 50, respectively.

[0090] First, the discrete cosine transform coefficient matrix of each sensor is arranged from the smallest to the largest into a sequence matrix C o It can be expressed as:

[0091]

[0092] The Shapiro-Wilk test (SW) was used to calculate the distribution test value as follows:

[0093]

[0094] Among them, when ξi>ξ(α), the coefficient is judged to obey the normal distribution, ξ(α) can be obtained by querying the Shapiro-Wilk test table, α=0.05, and n is the number of coefficients for SW test.

[0095] is the Shapiro-Wilk test table coefficient, μ i is the coefficient c ij The average value is calculated as follows:

[0096]

[0097] The Kolmogorov-Smirnov (KS) distribution test value is calculated as follows:

[0098]

[0099] in, When , the coefficient is judged to be subject to normal distribution. It can be obtained from the Kolmogorov-Smirnov test table, α = 0.05, where F(c ij ) is c ij The cumulative probability density function of for The cumulative frequency of .

[0100] F(c ij )and The calculation formula is as follows:

[0101]

[0102] In the formula, It's about c ij The probability density function of can be expressed as:

[0103]

[0104] Among them, σ i is the standard deviation of the discrete cosine transform coefficient of sensor i, which is calculated as follows:

[0105]

[0106]

[0107] Where j is the subscript sequence of any sensor i, and n is the total number of discrete cosine transform coefficients.

[0108] The 3-ξ criterion is used to test the normal distribution of the coefficients and then screen outliers. The calculation method is as follows:

[0109] P{|c ij -μ i |>(ξ*σ i )}=P(ξ); (11)

[0110] Among them, P(ξ=3)=0.9973, if |c ij -μ i |greater than ξ*σ i , then the corresponding probability can be calculated using formula (7). The larger the value of ξ, the stricter the threshold for screening outliers.

[0111] S24, outlier coefficient extraction: For different ξ values, the number of extracted outlier coefficients is different, but the total number of coefficients is the same. Generally speaking, the normal point coefficients account for a larger proportion of the total coefficients, and the outlier coefficients account for a smaller proportion of the total, but have a greater impact on the accuracy of the inverse discrete cosine transform calculation. During the encoding process, the normal point coefficients and the outlier coefficients are quantized and stored separately. When encoding the normal point coefficients, the normal point coefficients of a specific number of bits are quantized. When encoding the outlier coefficients, in addition to the quantization, the specific values ​​of the outlier coefficients and the order corresponding to each outlier coefficient are recorded and stored separately.

[0112] Step S3: quantize the outlier signal coefficient of each vibration sensor with an appropriate number of bits, then fuse the quantized normal signal coefficient and the outlier signal coefficient in bytes, and perform lossless compression on the fused coefficient.

[0113] Step S3 specifically includes:

[0114] S31, Discrete Cosine Transform Coefficient Bit Quantization Bit Number: The original mechanical vibration data is in floating point format and consists of 4 bytes. Discrete Cosine Transform Coefficient Bit Quantization is used to reduce the amount of data in the data transmission stage. Using fewer quantization bits helps to reduce the amount of data, but the quantization accuracy will also be reduced. Therefore, it is necessary to select a suitable quantization bit. In addition, since the outlier coefficient has a greater impact on the accuracy of the inverse discrete cosine transform, more bits are required for quantization. The discrete cosine transform coefficient quantization process is calculated as follows:

[0115]

[0116] Among them, χ max and χ min are the maximum and minimum values ​​of the interval to be mapped for each coefficient. i max and c i min is the maximum and minimum values ​​of the coefficients to be quantized, rounded to integers, is the quantization coefficient of any sensor, χ max and χmi n The calculation is as follows:

[0117] χ max =2 Q -1, χ min =0; (13)

[0118] Among them, Q is the number of bits of coefficient quantization, and the inverse quantization change can be expressed as:

[0119]

[0120] in, is the inverse quantization fusion coefficient, c imax and c imin They are the maximum and minimum values ​​of the quantization coefficient mapping interval respectively. and are the maximum and minimum values ​​of the quantization coefficients respectively.

[0121] S32, Discrete Cosine Transform Coefficient Fusion: Byte fusion is used to mix the coefficients of multiple sensors into one file, which can reduce the number of file headers. Coefficient fusion includes outlier fusion and normal point fusion.

[0122]

[0123] in, In the fusion file The probability of H is the information entropy of the fusion file. For m sensors, n points are taken for discrete cosine transform. After the transformation, there are still n coefficients. After filtering each sensor coefficient, the outlier coefficients are fused with the normal point coefficients. The normal point coefficient set is expressed as

[0124] S={s|(η,κ,...,v)};(16)

[0125] The outlier coefficient set is o=ns∈S, and the fusion sequence of normal point coefficients is and the fusion sequence of the outlier coefficient It can be expressed as:

[0126]

[0127]

[0128] where η,κ,v∈S≤n, and are the normal point quantization coefficient and the outlier point quantization coefficient.

[0129] Bit fusion is used to remove the information redundancy of the normal point and outlier fusion quantization coefficient sequence. First, the fusion sequence and Each coefficient is extracted as an 8-bit unsigned integer byte, each byte represents only 0 or 1, calculated as follows:

[0130]

[0131] in is the byte to be extracted, ">>" is the symbol for right shift, "&" is the symbol for AND operation, k is the number of right shifts, and q is the quantization bit.

[0132] After extracting the fused sequence as bytes, bit fusion is performed as follows:

[0133]

[0134] in is a new byte fused with the first 8 bits, “<<” is the symbol for left shift, “|” is the symbol for the “or” operation, and k is the number of left shifts.

[0135] Step S4: The compressed data is transmitted from the edge computing device to the twin virtual side for decoding, unpacking, fusion and quantization inversion, thereby realizing data reconstruction of the twin virtual side of the mechanical equipment.

[0136] Step S4 specifically includes:

[0137] S41, data decoding: read the header information of the decompressed data file and analyze the header data, including the vibration point, quantization bit number and sensor number. The detailed data structure is shown in Table 1.

[0138] Table 1 Data structure

[0139]

[0140]

[0141] S42, signal reconstruction: The stage of reducing data volume and information redundancy mainly includes coefficient quantization, bit fusion and lossless compression. The number of bits of the entire data compression is calculated as follows:

[0142]

[0143] Among them, N sv，i is the N of sensor i sv The specific value of B rv and B ov B r and B o The specific value of N ri and N oi are respectively N of sensor i r and N o The specific value of .

[0144] The reconstructed signal is calculated as follows

[0145]

[0146] in, is the inverse quantization fusion coefficient, is the inverse discrete cosine transformed vibration data.

[0147] The quality value of the reconstructed signal indicates the degree of reduction of the decoded data. The reconstructed signal quality is calculated using the root mean square difference percentage (PRD), the reconstructed signal-to-noise ratio (RSNR) and the data compression error (DCE) as follows:

[0148]

[0149]

[0150]

[0151] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention, and any reference numerals in the claims should not be regarded as limiting the claims involved.

[0152] In addition, it should be understood that although this specification is described in terms of implementation methods, not every implementation method contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

Claims

1. A digital twin driving model of electromechanical equipment status based on multi-sensor monitoring, characterized in that: The method used includes the following steps: Step S1, using a vibration sensor to collect original vibration signal data of mechanical equipment, and transmitting the original vibration signal data to an edge computing device for encoding; Step S2, performing discrete cosine transform on the encoded original vibration signal data, and clustering the normal point signal coefficients and outlier signal coefficients obtained after the discrete cosine transform using the 3-ξ criterion; Step S3, quantizing the normal point signal coefficient and the outlier signal coefficient corresponding to each vibration sensor using a selected number of bits, then fusing the quantized normal point signal coefficient and the outlier signal coefficient in bytes, and performing lossless compression on the obtained fused coefficients; Step S4, transmits the compressed data from the edge computing device to the twin virtual side for decoding, unpacking, fusion and quantization inversion to obtain a reconstructed signal, thereby realizing data reconstruction of the twin virtual side of the mechanical equipment.

2. According to claim 1, the digital twin driving model of electromechanical equipment status based on multi-sensor monitoring is characterized in that: Step S1 specifically includes: Multiple vibration sensors are placed around the planetary gearbox of the mechanical equipment to collect the original vibration signal data of the mechanical equipment. Each vibration sensor is responsible for collecting data from a vibration test point. The data sampling card is connected to the edge computing device for fusion encoding of the vibration signal.

3. The digital twin driving model of electromechanical equipment status based on multi-sensor monitoring according to claim 1 is characterized in that: Step S2 specifically includes the following steps: S21, define the original vibration signal data X: Where m is the number of vibration sensors on the mechanical equipment, and n is the number of sampling points of each vibration sensor in a period of time; Perform discrete cosine transform on the original vibration signal data X to obtain the discrete cosine transform coefficient matrix C: Among them, the discrete cosine transform coefficient S22, discrete cosine transform coefficients represent the degree of signal energy concentration. Larger amplitudes are close to zero, and smaller amplitudes are far from zero. Discrete cosine transform coefficients with amplitudes greater than a set value and amplitudes less than a set value are defined as outlier coefficients and normal point coefficients, respectively. S23, outlier coefficient detection: First, the elements in the discrete cosine transform coefficient matrix of each vibration sensor are arranged from the smallest to the largest into a sequence matrix C o : Represents the sequence matrix C o The Shapiro-Wilk test method is used to calculate the distribution test value ξ i : Among them, ξ i >ξ(α), the distribution test value ξ i It is determined to be normally distributed, ξ(α) is obtained by querying the Shapiro-Wilk test table, α = 0.05, n is the number of discrete cosine transform coefficients, is the Shapiro-Wilk test table coefficient, μ i is the discrete cosine transform coefficient c ij The average value of: Calculate the distribution test value using Kolmogorov-Smirnov in, When , the distribution test value It is judged to be subject to normal distribution. From the Kolmogorov-Smirnov test table, we can find that α = 0.05, where F(c ij ) is c ij The cumulative probability density function of for The cumulative frequency of F(c ij )and The calculation formula is as follows: Formula 7: In the formula, It's about c ij The probability density function of is: Among them, σ i is the standard deviation of the discrete cosine transform coefficient of the i-th vibration sensor: Where, j is the subscript sequence of the i-th vibration sensor; The 3-ξ criterion is used to test the normal distribution of discrete cosine transform coefficients and then screen outliers: P{|c i -m i |>(ξ*σ i )}=P(ξ); Among them, P(ξ=3)=0.9973, if |c ij -μ i |greater than ξ*σ i , then P{|c ij -μ i |>(ξ*σ i )} can be calculated by formula 7; the larger the value of ξ, the larger the threshold for screening outliers; S24, in the encoding process of the original vibration signal data, the normal point coefficients and the outlier coefficients are quantized and stored respectively. When encoding the normal point coefficients, the normal point coefficients of specific bits are quantized. When encoding the outlier coefficients, in addition to quantizing the specific values ​​of the outlier coefficients, the order corresponding to each outlier coefficient is also recorded and stored separately.

4. The digital twin driving model of electromechanical equipment status based on multi-sensor monitoring according to claim 1 is characterized in that: Step S3 specifically includes the following steps: S31, Discrete cosine transform coefficient bit quantization number: Among them, χ max and χ min are the maximum and minimum values ​​of the interval to be mapped for each discrete cosine transform coefficient. imax and c imin is the maximum and minimum value of the coefficient to be quantized, is the quantization coefficient of any sensor, χ max and χ min The calculation is as follows: x max =2 Q -1,x min =0; Among them, Q is the number of bits of discrete cosine transform coefficient quantization, and the inverse quantization change is expressed as: in, is the inverse quantization fusion coefficient; and are the maximum and minimum values ​​of the quantization coefficients respectively; S32, Discrete Cosine Transform Coefficient Fusion: Byte fusion is used to mix the discrete cosine transform coefficients of multiple vibration sensors into a fusion file; discrete cosine transform coefficient fusion includes outlier coefficient fusion and normal point coefficient fusion: in, In the fusion file The probability of, H is the information entropy of the fusion file; for m vibration sensors, n points are taken for discrete cosine transform, and there are still n coefficients after the transformation. After filtering the discrete cosine transform coefficients of each sensor, the outlier coefficients are fused with the normal point coefficients respectively; Fusion sequence of normal point coefficients and the fusion sequence of the outlier coefficient It is expressed as: Among them, the set of normal point coefficients is S = {s|(η,κ,...,v)}; the set of outlier coefficients is ο = ns∈S, η,κ,v∈S≤n, and They represent the normal point quantization coefficient and the outlier point quantization coefficient respectively; Bit fusion is used to remove the information redundancy of the fusion sequence of normal point quantization coefficients and outlier point quantization coefficients. First, the fusion sequence and Each coefficient is extracted as an 8-bit unsigned integer byte, each byte represents only 0 or 1: in is the byte to be extracted, ">>" is the symbol for right shift, "&" is the symbol for AND operation, k is the number of right shifts, and q is the quantization bit; After extracting the fused sequence as bytes, bit fusion is performed as follows: in is the new byte fused with the first 8 bits, "<<" is the symbol of left shift, "|" is the symbol of "or" operation, and k is the number of left shifts.

5. The digital twin driving model of electromechanical equipment status based on multi-sensor monitoring according to claim 1 is characterized in that: Step S4 specifically includes the following steps: S41, data decoding: decompressing the received compressed data and reading the header information of the decompressed data file; S42, signal reconstruction: including coefficient quantization, bit fusion and lossless compression; the number of data compression bits B is: Among them, N sv,i is the specific value of the number of bits of the i-th sensor, B rv and B ov are the specific values ​​of the quantization bit number of the normal point discrete cosine transform coefficient and the quantization bit number of the outlier point discrete cosine transform coefficient respectively; N ri and N oi are the specific values ​​of the number of fused unquantized normal point discrete cosine transform coefficients and the number of fused unquantized outlier discrete cosine transform coefficients of the i-th sensor respectively; The reconstructed signal is calculated as follows: in, is the inverse quantization fusion coefficient, is the inverse discrete cosine transformed vibration data.

Images